</>MCP Agents Market
Plugin

pro-workflow

by rohitg002.8kJavaScriptUpdated 2026-08-31

Claude Code learns from your corrections: self-correcting memory that compounds over 50+ sessions. Context engineering, parallel worktrees, agent teams, and 17 battle-tested skills.

Claude CodeCursorWindsurfCodexGemini CLIOpenCodeGitHub CopilotGoose

Pro Workflow is a Claude Code plugin and cross-agent skill bundle that gives AI coding assistants persistent memory and learning capabilities across sessions. It captures corrections as searchable rules stored in SQLite with FTS5 indexing, so the AI stops repeating the same mistakes after being corrected once. The plugin adds 41 skills, 8 specialized agents, 23 slash commands, and 37 event hooks that enable persistent research wikis, multi-phase development workflows with validation gates, parallel worktrees, quality gates, and budget-tracked auto-research loops. After 50+ sessions, correction rates approach zero as the AI compounds learnings and builds domain-specific knowledge bases.

Key Features

Self-correcting memory that persists corrections as FTS5-searchable rules in SQLite, auto-loaded in every session
Persistent research wikis with auto-research loops (budget-capped BFS over web/arXiv/GitHub) that grow knowledge bases overnight
Hybrid retrieval combining BM25 and vector search with reciprocal rank fusion for wiki queries
Multi-LLM council deliberation across Anthropic/OpenAI/OpenRouter/Fireworks for high-stakes decisions
37 hook scripts across 24 events (SessionStart, UserPromptSubmit, PreToolUse, PreCompact, etc.) for intelligent automation
Quality gates including LLM-powered commit validation, secret scanning, and blast-radius checks before destructive operations
Parallel worktrees and agent teams for zero-downtime multi-session orchestration
Cross-agent support for Claude Code, Cursor, Windsurf, and 29+ other AI coding assistants via skills add

Use Cases

  • 01Eliminate repetitive corrections by capturing coding conventions, test patterns, and project-specific rules once
  • 02Build persistent domain knowledge wikis (frameworks, APIs, architecture decisions) that grow via auto-research and are injected when relevant
  • 03Run multi-phase feature development (Research → Plan → Implement → Review) with validation gates between phases
  • 04Prevent secrets and destructive operations from reaching production with LLM-powered pre-commit hooks
  • 05Coordinate multiple AI agents in parallel worktrees for large-scale refactors without context conflicts
  • 06Track session costs, token efficiency, and correction trends with built-in analytics and budget alerts

Related Plugins

View more

pro-workflow — FAQ

What is pro-workflow?+

Pro Workflow is a Claude Code plugin and skill bundle that adds persistent SQLite-backed memory to AI coding assistants. It captures corrections as searchable rules, maintains FTS5-indexed research wikis, and provides orchestration tools like quality gates, parallel worktrees, and multi-LLM councils. It works natively in Claude Code and installs as skills in 32+ other AI agents.

How do I install pro-workflow in Claude Code?+

In Claude Code, run `/plugin marketplace add rohitg00/pro-workflow` then `/plugin install pro-workflow@pro-workflow`. For other agents like Cursor or Windsurf, use `npx skills add rohitg00/pro-workflow` followed by `skills sync` to register the skills.

Which AI coding assistants does pro-workflow support?+

It supports Claude Code natively as a plugin, plus Cursor, Codex, Gemini CLI, Windsurf, OpenCode, GitHub Copilot CLI, Goose, Roo, and 27+ other agents via skills add installation.

Do I need API keys to use pro-workflow?+

Core features (self-correction memory, quality gates, hooks) require no API keys. Optional features require keys: hybrid wiki retrieval needs OPENAI_API_KEY or VOYAGE_API_KEY for embeddings, and the multi-LLM council needs ANTHROPIC_API_KEY, OPENAI_API_KEY, OPENROUTER_API_KEY, or FIREWORKS_API_KEY.

Is pro-workflow free?+

Yes, pro-workflow is MIT-licensed and free to use. Optional features like auto-research loops and multi-LLM councils consume tokens from your own API keys, which you can control with budget caps (e.g. `--budget-usd 0.50`).

How does the self-correction memory work?+

When you correct Claude (e.g., 'Don't mock the database in tests'), run `/learn-rule` to capture it. The correction is stored in SQLite with FTS5 indexing and auto-loaded on every SessionStart, so Claude never repeats that mistake. After 50 sessions, correction rates approach zero.

How do I install pro-workflow?+

Open the source repository on GitHub and follow its README. pro-workflow is a plugin — MCP Agents Market links you directly to the official repo.

Is pro-workflow free?+

pro-workflow is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches